steelman
Skill0 starsChanged 3 months agoMITView on GitHub
对一个想法、结论或方案做双向钢人(steelman)压力测试:不偏向用户也不为反对而反对,把正反两边做成最强版后仍给出判断。用户说压力测试我的结论、钢人、steelman、帮我找反方、唱反调、站在我反面想想、我是不是想错了时使用;立场不清晰时改用 grilling。
13 other projects use a version of this.
Adopted by
Most recently taken up by
- rahul-kulkarni105/skillstheir own version, first seen Nov 8, 2026
- wangjs-jacky/jacky-skillstheir own version, first seen Nov 8, 2026
- Bobby-cell-commits/steelman-skilltheir own version, first seen Nov 1, 2026
- natexai2026/2030-skillstheir own version, first seen Oct 20, 2026
- malinskibeniamin/skillstheir own version, first seen Oct 17, 2026
- rockcookies/skillstheir own version, first seen Oct 13, 2026
Other versions13 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 13
- wangjs-jacky/jacky-skills★ 15same name
- techiejd/claude-skill-steelman★ 10same name
- Bobby-cell-commits/steelman-skill★ 9same name
- damionrashford/Adversaria★ 9same name
- dgriffith/combat-epistemology★ 7same name
- malinskibeniamin/skills★ 4same name
- MichaelLeeHobbs/skills★ 1same name
- rockcookies/skills★ 1same name
- dilulloriccardo-arch/rigor-skills★ 0same name
- Nankov-ai/anthropics-skills★ 0same name
3 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository19 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- global/CLAUDE.mdCLAUDE.md
- article-ingestSkill
- git-workflowSkill
- grillingSkill
- handoffSkill
- hv-analysisSkill
- leaderSkill
- lqSkill
- mmd-explainSkill
- neat-freakSkill
- pickupSkill
- playgroundSkill
- session-logSkill
- session-searchSkill
- taskboardSkill
- ui-optionsSkill
- zettel-distillSkill
The file
Files
steelman — 双向钢人压力测试
反谄媚机制:不顺着用户说,也不为反对而反对——把正反两边都做成最强版,让分歧落在可检验的变量上,最后仍要给出自己的判断。流程改造自数字生命卡兹克的「双向钢人」Prompt(vault 卡 REF-双向钢人Prompt-数字生命卡兹克),第 1 步来自 Rapoport 法则(Dennett《直觉泵和其他思考工具》第 3 章)。
流程(顺序执行,不跳步;只在第 4 步停下来等回答)
- 复述:用自己的话复述用户的问题、立场和它依赖的关键前提,标明「以下按这个理解往下做」,然后直接继续。用户不打断即视为复述准确,打断了就按纠正重跑受影响的步骤。这一步防的是「让 AI 反驳我,它去找莫名其妙的点」——打的不是用户真正持有、真正依赖的东西(替对方造一个他并不持有的「最强论证」也是同一种错,等于反向稻草人)——所以复述必须显式写出来让用户能一眼核对,但不拿它当闸门卡住流程:复述错了是单点可改,先做一版靠反馈修正,比每次多一个来回便宜。
- 双向最强:分别给正方(用户立场)和反方的最强论证。反方只朝用户会认可的前提方向强化,不引入用户明确否定过的前提;两边都注明「这个论证最依赖哪个假设」。
- 关键变量:指出最可能改变结论的 1~2 个分歧点,以及什么证据或实验能分辨它们(double crux:双方各说什么证据会让自己改主意,找那个交集)。
- 只问一题:向用户提一个最关键的必答问题——答案直接决定结论倒向哪边——然后停下来等回答。这里不替用户假设:第 1 步的复述是 agent 能做、用户只需核对的事,所以不阻塞;这一题是只有用户能答、agent 怎么假设都是替她拍板的事,所以必须等。
- 判断:拿到回答后给出自己的判断与行动建议,标把握几成。不骑墙:两边论证都摆完后仍要说自己站哪边、为什么。
边界
- 与 grilling 的分工:grilling 把还没成形的想法问清楚(向用户提问为主);本 skill 压力测试一个已成形的立场(产出论证为主)。第 1 步复述时发现立场、前提或目标说不清(复述不出一个可被反驳的命题),先触发 grilling 问到立场明确,再回到本流程从第 1 步开始——不要对着一个模糊的立场造钢人,那样正反两边都是空的。
- 用户中途纠正复述或给出新事实时,回到受影响的最早一步重跑那部分,不推倒重来。
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

